Computer Networks, Spring 2013 Socket Programming Computer Networks, Spring 2013

Lecture Today Motivation for sockets What’s in a socket? Working with socket Concurrent network applications Project 1

Why Socket? How can I program a network application? Share data Send messages Finish course projects... IPC - Interprocess Communication

Network Layering Application Application 7 Presentation Presentation 6 Session Session 5 Transport Transport 4 Network Network Network 3 Data link Data link Data link 2 Physical Physical Physical 1

Network Layering Why layering? Application Application 7 Presentation 6 Session Session 5 Transport Transport 4 Network Network Network 3 Data link Data link Data link 2 Physical Physical Physical 1

Layering Makes it Easier Application programmer Doesn’t need to send IP packets Doesn’t need to send Ethernet frames Doesn’t need to know how TCP implements reliability Only need a way to pass the data down Socket is the API to access transport layer functions

What Lower Layer Need to Know? We pass the data down. What else does the lower layer need to know?

What Lower Layer Need to Know? We pass the data down. What else does the lower layer need to know? How to identify the destination process? Where to send the data? (Addressing) What process gets the data when it is there? (Multiplexing)

Identify the Destination Addressing IP address hostname (resolve to IP address via DNS) Multiplexing port Server socket address 208.216.181.15:80 Client socket address 128.2.194.242:3479 FTP Server (port 21) Client HTTP Server (port 80) Connection socket pair (128.2.194.242:3479, 208.216.181.15:80) Client host address 128.2.194.242 Server host address 208.216.181.15

Sockets How to use sockets Setup socket Send and Receive Where is the remote machine (IP address, hostname) What service gets the data (port) Send and Receive Designed just like any other I/O in unix send -- write recv -- read Close the socket

Overview Client Server open_listenfd open_clientfd Connection request socket socket bind open_listenfd open_clientfd listen Connection request connect accept read close EOF read write Client / Server Session

Step 1 – Setup Socket Both client and server need to setup the socket int socket(int domain, int type, int protocol); domain AF_INET -- IPv4 (AF_INET6 for IPv6) type SOCK_STREAM -- TCP SOCK_DGRAM -- UDP protocol For example, int sockfd = socket(AF_INET, SOCK_STREAM, 0);

Step 2 (Server) - Binding Only server need to bind int bind(int sockfd, const struct sockaddr *my_addr, socklen_t addrlen); sockfd file descriptor socket() returned my_addr struct sockaddr_in for IPv4 cast (struct sockaddr_in*) to (struct sockaddr*) struct sockaddr_in { short sin_family; // e.g. AF_INET unsigned short sin_port; // e.g. htons(3490) struct in_addr sin_addr; // see struct in_addr, below char sin_zero[8]; // zero this if you want to }; struct in_addr { unsigned long s_addr; // load with inet_aton()

What is that Cast? bind() takes in protocol-independent (struct sockaddr*) C’s polymorphism There are structs for IPv6, etc. struct sockaddr { unsigned short sa_family; // address family char sa_data[14]; // protocol address };

Step 2 (Server) - Binding contd. addrlen size of the sockaddr_in struct sockaddr_in saddr; int sockfd; unsigned short port = 80; if((sockfd=socket(AF_INET, SOCK_STREAM, 0) < 0) { // from back a couple slides printf(“Error creating socket\n”); ... } memset(&saddr, '\0', sizeof(saddr)); // zero structure out saddr.sin_family = AF_INET; // match the socket() call saddr.sin_addr.s_addr = htonl(INADDR_ANY); // bind to any local address saddr.sin_port = htons(port); // specify port to listen on if((bind(sockfd, (struct sockaddr *) &saddr, sizeof(saddr)) < 0) { // bind! printf(“Error binding\n”);

What is htonl(), htons()? Byte ordering Network order is big-endian Host order can be big- or little-endian x86 is little-endian SPARC is big-endian Conversion htons(), htonl(): host to network short/long ntohs(), ntohl(): network order to host short/long What need to be converted? Addresses Port etc.

Step 3 (Server) - Listen Now we can listen sockfd backlog For example, int listen(int sockfd, int backlog); sockfd again, file descriptor socket() returned backlog number of pending connections to queue For example, listen(sockfd, 5);

Step 4 (Server) - Accept Server must explicitly accept incoming connections int accept(int sockfd, struct sockaddr *addr, socklen_t *addrlen) sockfd again... file descriptor socket() returned addr pointer to store client address, (struct sockaddr_in *) cast to (struct sockaddr *) addrlen pointer to store the returned size of addr, should be sizeof(*addr) For example int isock=accept(sockfd, (struct sockaddr_in *) &caddr, &clen);

Put Server Together struct sockaddr_in saddr, caddr; int sockfd, clen, isock; unsigned short port = 80; if((sockfd=socket(AF_INET, SOCK_STREAM, 0) < 0) { // from back a couple slides printf(“Error creating socket\n”); ... } memset(&saddr, '\0', sizeof(saddr)); // zero structure out saddr.sin_family = AF_INET; // match the socket() call saddr.sin_addr.s_addr = htonl(INADDR_ANY); // bind to any local address saddr.sin_port = htons(port); // specify port to listen on if((bind(sockfd, (struct sockaddr *) &saddr, sizeof(saddr)) < 0) { // bind! printf(“Error binding\n”); if(listen(sockfd, 5) < 0) { // listen for incoming connections printf(“Error listening\n”); clen=sizeof(caddr) if((isock=accept(sockfd, (struct sockaddr *) &caddr, &clen)) < 0) { // accept one printf(“Error accepting\n”);

What about client? Client need not bind, listen, and accept All client need to do is to connect int connect(int sockfd, const struct sockaddr *saddr, socklen_t addrlen); For example, connect(sockfd, (struct sockaddr *) &saddr, sizeof(saddr));

Domain Name System (DNS) What if I want to send data to “www.slashdot.org”? DNS: Conceptually, DNS is a database collection of host entries hostname -> IP address struct hostent *gethostbyname(const char *name); IP address -> hostname struct hostent *gethostbyaddr(const char *addr, int len, int type); struct hostent { char *h_name; // official hostname char **h_aliases; // vector of alternative hostnames int h_addrtype; // address type, e.g. AF_INET int h_length; // length of address in bytes, e.g. 4 for IPv4 char **h_addr_list; // vector of addresses char *h_addr; // first host address, synonym for h_addr_list[0] };

Put Client Together struct sockaddr_in saddr; struct hostent *h; int sockfd, connfd; unsigned short port = 80; if((sockfd=socket(AF_INET, SOCK_STREAM, 0) < 0) { // from back a couple slides printf(“Error creating socket\n”); ... } if((h=gethostbyname(“www.slashdot.org”)) == NULL) { // Lookup the hostname printf(“Unknown host\n”); memset(&saddr, '\0', sizeof(saddr)); // zero structure out saddr.sin_family = AF_INET; // match the socket() call memcpy((char *) &saddr.sin_addr.s_addr, h->h_addr_list[0], h->h_length); // copy the address saddr.sin_port = htons(port); // specify port to connect to if((connfd=connect(sockfd, (struct sockaddr *) &saddr, sizeof(saddr)) < 0) { // connect! printf(“Cannot connect\n”);

We Are Connected Server accepting connections and client connecting to servers Send and receive data ssize_t read(int fd, void *buf, size_t len); ssize_t write(int fd, const void *buf, size_t len); For example, read(sockfd, buffer, sizeof(buffer)); write(sockfd, “hey\n”, strlen(“hey\n”));

TCP Framing TCP does NOT guarantee message boundaries IRC commands are terminated by a newline But you may not get one at the end of read(), e.g. One Send “Hello\n” Multiple Receives “He”, “llo\n” If you don’t get the entire line from one read(), use a buffer

Revisited Client Server open_listenfd open_clientfd Connection request socket socket bind open_listenfd open_clientfd listen Connection request connect accept read close EOF read write Client / Server Session

Close the Socket Don’t forget to close the socket descriptor, like a file int close(int sockfd); Now server can loop around and accept a new connection when the old one finishes What’s wrong here?

Server Flaw client 1 server client 2 Server blocks waiting for call accept call connect ret connect ret accept call fgets call read Server blocks waiting for data from Client 1 call connect User goes out to lunch Client 1 blocks waiting for user to type in data Client 2 blocks waiting to complete its connection request until after lunch! Taken from D. Murray, R. Bryant, and G. Langale 15-441/213 slides

Concurrent Servers client 1 server client 2 User goes out to lunch call accept call connect call connect ret connect ret accept call fgets call read (don’t block) call accept User goes out to lunch Client 1 blocks waiting for user to type in data ret connect call fgets ret accept write call read call read write end read close close Taken from D. Murray, R. Bryant, and G. Langale 15-441/213 slides

Concurrency Threading Select() Easier to understand Race conditions increase complexity Select() Explicit control flows, no race conditions Explicit control more complicated There is no clear winner, but you MUST use select()…

What is select()? Monitor multiple descriptors How does it work? Setup sets of sockets to monitor select(): blocking until something happens “Something” could be Incoming connection: accept() Clients sending data: read() Pending data to send: write() Timeout

Concurrency – Step 1 Allowing address reuse Then we set the sockets to be non-blocking int sock, opts=1; sock = socket(...); // To give you an idea of where the new code goes setsockopt(sock, SOL_SOCKET, SO_REUSEADDR, &opts, sizeof(opts)); if((opts = fcntl(sock, F_GETFL)) < 0) { // Get current options printf(“Error...\n”); ... } opts = (opts | O_NONBLOCK); // Don't clobber your old settings if(fcntl(sock, F_SETFL, opts) < 0) { bind(...); // To again give you an idea where the new code goes

Concurrency – Step 2 Monitor sockets with select() maxfd int select(int maxfd, fd_set *readfds, fd_set *writefds, fd_set *exceptfds, const struct timespec *timeout); maxfd max file descriptor + 1 fd_set: bit vector with FD_SETSIZE bits readfds: bit vector of read descriptors to monitor writefds: bit vector of write descriptors to monitor exceptfds: set to NULL timeout how long to wait without activity before returning

What about bit vectors? void FD_ZERO(fd_set *fdset); clear out all bits void FD_SET(int fd, fd_set *fdset); set one bit void FD_CLR(int fd, fd_set *fdset); clear one bit int FD_ISSET(int fd, fd_set *fdset); test whether fd bit is set

The Server // socket() call and non-blocking code is above this point if((bind(sockfd, (struct sockaddr *) &saddr, sizeof(saddr)) < 0) { // bind! printf(“Error binding\n”); ... } if(listen(sockfd, 5) < 0) { // listen for incoming connections printf(“Error listening\n”); clen=sizeof(caddr); // Setup pool.read_set with an FD_ZERO() and FD_SET() for // your server socket file descriptor. (whatever socket() returned) while(1) { pool.ready_set = pool.read_set; // Save the current state pool.nready = select(pool.maxfd+1, &pool.ready_set, &pool.write_set, NULL, NULL); if(FD_ISSET(sockfd, &pool.ready_set)) { // Check if there is an incoming conn isock=accept(sockfd, (struct sockaddr *) &caddr, &clen); // accept it add_client(isock, &pool); // add the client by the incoming socket fd check_clients(&pool); // check if any data needs to be sent/received from clients close(sockfd);

What is pool? typedef struct { /* represents a pool of connected descriptors */ int maxfd; /* largest descriptor in read_set */ fd_set read_set; /* set of all active read descriptors */ fd_set write_set; /* set of all active read descriptors */ fd_set ready_set; /* subset of descriptors ready for reading */ int nready; /* number of ready descriptors from select */ int maxi; /* highwater index into client array */ int clientfd[FD_SETSIZE]; /* set of active descriptors */ rio_t clientrio[FD_SETSIZE]; /* set of active read buffers */ ... // ADD WHAT WOULD BE HELPFUL FOR PROJECT1 } pool;

What about checking clients? The main loop only tests for incoming connections There are other reasons the server wakes up Clients are sending data, pending data to write to buffer, clients closing connections, etc. Store all client file descriptors in pool Keep the while(1) loop thin Delegate to functions Come up with your own design

Summary Sockets socket setup I/O close Client: socket()----------------------->connect()->I/O->close() Server: socket()->bind()->listen()->accept()--->I/O->close() DNS gethostbyname() Concurrency select() Bit vector operations fd_set, FD_ZERO(), FD_SET(), FD_CLR(), FD_ISSET()

About Project 1 Standalone IRC server Checkpoint 1: subversion and Makefile Check in a Makefile and source code Makefile can build executable named sircd No server functions necessary Checkpoint 2: echo server Use select() to handle multiple clients

Suggestions Start early! Read the man pages Work ahead of checkpoints Read the man pages Email (xil at cs dot cmu dot edu) if you didn’t get a svn username and password